From 8fd57a97f203edac3f7e8681eafe752294386a24 Mon Sep 17 00:00:00 2001 From: Benoit Steiner Date: Tue, 7 Jun 2016 18:22:18 -0700 Subject: Enable the vectorization of adds and mults of fp16 --- Eigen/src/Core/arch/CUDA/PacketMathHalf.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'Eigen/src/Core/arch') diff --git a/Eigen/src/Core/arch/CUDA/PacketMathHalf.h b/Eigen/src/Core/arch/CUDA/PacketMathHalf.h index 51386506f..959dff886 100644 --- a/Eigen/src/Core/arch/CUDA/PacketMathHalf.h +++ b/Eigen/src/Core/arch/CUDA/PacketMathHalf.h @@ -28,6 +28,8 @@ template<> struct packet_traits : default_packet_traits AlignedOnScalar = 1, size=2, HasHalfPacket = 0, + HasAdd = 1, + HasMul = 1, HasDiv = 1, HasSqrt = 1, HasRsqrt = 1, -- cgit v1.2.3